There is no such thing… it is not algorithmically impossible, but it is very very difficult. There is software that is extremely expensive (Melodyne) that, with a lot of work, can do a really terrible job of it – but you would be doing most of the work, not the software… and chances are, you wouldn’t be willing to play the results to a crowd.
Such software will eventually exist, but it’s going to take us a while to get there and chances are, when we do, it will be an expensive commercial product for the first long while (and will probably be produced by Celemony).
I have experimented with this a lot, and have some interesting angles I’m working, but at this point I know enough to know that it will be a while before anyone can actually pull this off in anything resembling an automated fashion.
.
